[PATCH v2 3/5] drm/xe/bo: Add a bo remove callback
Thomas Hellström
thomas.hellstrom at linux.intel.com
Mon Mar 24 15:48:56 UTC 2025
On Mon, 2025-03-24 at 14:41 +0000, Matthew Auld wrote:
> On 21/03/2025 16:34, Thomas Hellström wrote:
> > On device unbind, migrate exported bos, including pagemap bos to
> > system. This allows importers to take proper action without
> > disruption. In particular, SVM clients on remote devices may
> > continue as if nothing happened, and can chose a different
> > placement.
> >
> > The evict_flags() placement is chosen in such a way that bos that
> > aren't exported are purged.
> >
> > For pinned bos, we unmap DMA, but their pages are not freed yet
> > since we can't be 100% sure they are not accessed.
> >
> > v2:
> > - Address review comments. (Matthew Auld).
> >
> > Signed-off-by: Thomas Hellström <thomas.hellstrom at linux.intel.com>
> > ---
> > drivers/gpu/drm/xe/xe_bo.c | 54 ++++++++++++++++++++---
> > -
> > drivers/gpu/drm/xe/xe_bo.h | 2 +
> > drivers/gpu/drm/xe/xe_bo_evict.c | 62
> > ++++++++++++++++++++++++++--
> > drivers/gpu/drm/xe/xe_bo_evict.h | 1 +
> > drivers/gpu/drm/xe/xe_device.c | 6 ++-
> > drivers/gpu/drm/xe/xe_device_types.h | 6 ++-
> > 6 files changed, 117 insertions(+), 14 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/xe/xe_bo.c
> > b/drivers/gpu/drm/xe/xe_bo.c
> > index 64f9c936eea0..9d043d2c30fd 100644
> > --- a/drivers/gpu/drm/xe/xe_bo.c
> > +++ b/drivers/gpu/drm/xe/xe_bo.c
> > @@ -55,6 +55,8 @@ static struct ttm_placement sys_placement = {
> > .placement = &sys_placement_flags,
> > };
> >
> > +static struct ttm_placement purge_placement;
> > +
> > static const struct ttm_place tt_placement_flags[] = {
> > {
> > .fpfn = 0,
> > @@ -281,6 +283,8 @@ int xe_bo_placement_for_flags(struct xe_device
> > *xe, struct xe_bo *bo,
> > static void xe_evict_flags(struct ttm_buffer_object *tbo,
> > struct ttm_placement *placement)
> > {
> > + struct xe_device *xe = container_of(tbo->bdev,
> > typeof(*xe), ttm);
> > + bool device_unplugged = drm_dev_is_unplugged(&xe->drm);
> > struct xe_bo *bo;
> >
> > if (!xe_bo_is_xe_bo(tbo)) {
> > @@ -290,7 +294,7 @@ static void xe_evict_flags(struct
> > ttm_buffer_object *tbo,
> > return;
> > }
> >
> > - *placement = sys_placement;
> > + *placement = device_unplugged ? purge_placement :
> > sys_placement;
> > return;
> > }
> >
> > @@ -300,6 +304,11 @@ static void xe_evict_flags(struct
> > ttm_buffer_object *tbo,
> > return;
> > }
> >
> > + if (device_unplugged && !tbo->base.dma_buf) {
> > + *placement = purge_placement;
> > + return;
> > + }
> > +
> > /*
> > * For xe, sg bos that are evicted to system just triggers
> > a
> > * rebind of the sg list upon subsequent validation to
> > XE_PL_TT.
> > @@ -657,11 +666,20 @@ static int xe_bo_move_dmabuf(struct
> > ttm_buffer_object *ttm_bo,
> > struct xe_ttm_tt *xe_tt = container_of(ttm_bo->ttm, struct
> > xe_ttm_tt,
> > ttm);
> > struct xe_device *xe = ttm_to_xe_device(ttm_bo->bdev);
> > + bool device_unplugged = drm_dev_is_unplugged(&xe->drm);
> > struct sg_table *sg;
> >
> > xe_assert(xe, attach);
> > xe_assert(xe, ttm_bo->ttm);
> >
> > + if (device_unplugged && new_res->mem_type == XE_PL_SYSTEM
> > &&
> > + ttm_bo->sg) {
> > + dma_resv_wait_timeout(ttm_bo->base.resv,
> > DMA_RESV_USAGE_BOOKKEEP,
> > + false,
> > MAX_SCHEDULE_TIMEOUT);
> > + dma_buf_unmap_attachment(attach, ttm_bo->sg,
> > DMA_BIDIRECTIONAL);
> > + ttm_bo->sg = NULL;
> > + }
> > +
> > if (new_res->mem_type == XE_PL_SYSTEM)
> > goto out;
> >
> > @@ -1224,6 +1242,31 @@ int xe_bo_restore_pinned(struct xe_bo *bo)
> > return ret;
> > }
> >
> > +int xe_bo_dma_unmap_pinned(struct xe_bo *bo)
> > +{
> > + struct ttm_buffer_object *ttm_bo = &bo->ttm;
> > + struct ttm_tt *tt = ttm_bo->ttm;
> > +
> > + if (tt) {
> > + struct xe_ttm_tt *xe_tt = container_of(tt,
> > typeof(*xe_tt), ttm);
> > +
> > + if (ttm_bo->type == ttm_bo_type_sg && ttm_bo->sg)
> > {
> > + dma_buf_unmap_attachment(ttm_bo-
> > >base.import_attach,
> > + ttm_bo->sg,
> > +
> > DMA_BIDIRECTIONAL);
> > + ttm_bo->sg = NULL;
> > + xe_tt->sg = NULL;
> > + } else if (xe_tt->sg) {
> > + dma_unmap_sgtable(xe_tt->xe->drm.dev,
> > xe_tt->sg,
> > + DMA_BIDIRECTIONAL, 0);
> > + sg_free_table(xe_tt->sg);
> > + xe_tt->sg = NULL;
> > + }
> > + }
> > +
> > + return 0;
> > +}
> > +
> > static unsigned long xe_ttm_io_mem_pfn(struct ttm_buffer_object
> > *ttm_bo,
> > unsigned long page_offset)
> > {
> > @@ -2102,12 +2145,9 @@ int xe_bo_pin_external(struct xe_bo *bo)
> > if (err)
> > return err;
> >
> > - if (xe_bo_is_vram(bo)) {
> > - spin_lock(&xe->pinned.lock);
> > - list_add_tail(&bo->pinned_link,
> > - &xe->pinned.external_vram);
> > - spin_unlock(&xe->pinned.lock);
> > - }
> > + spin_lock(&xe->pinned.lock);
> > + list_add_tail(&bo->pinned_link, &xe-
> > >pinned.external);
> > + spin_unlock(&xe->pinned.lock);
> > }
> >
> > ttm_bo_pin(&bo->ttm);
> > diff --git a/drivers/gpu/drm/xe/xe_bo.h
> > b/drivers/gpu/drm/xe/xe_bo.h
> > index ec3e4446d027..05479240bf75 100644
> > --- a/drivers/gpu/drm/xe/xe_bo.h
> > +++ b/drivers/gpu/drm/xe/xe_bo.h
> > @@ -276,6 +276,8 @@ int xe_bo_evict(struct xe_bo *bo, bool
> > force_alloc);
> > int xe_bo_evict_pinned(struct xe_bo *bo);
> > int xe_bo_restore_pinned(struct xe_bo *bo);
> >
> > +int xe_bo_dma_unmap_pinned(struct xe_bo *bo);
> > +
> > extern const struct ttm_device_funcs xe_ttm_funcs;
> > extern const char *const xe_mem_type_to_name[];
> >
> > diff --git a/drivers/gpu/drm/xe/xe_bo_evict.c
> > b/drivers/gpu/drm/xe/xe_bo_evict.c
> > index 1eeb3910450b..ba0260900868 100644
> > --- a/drivers/gpu/drm/xe/xe_bo_evict.c
> > +++ b/drivers/gpu/drm/xe/xe_bo_evict.c
> > @@ -93,8 +93,8 @@ int xe_bo_evict_all(struct xe_device *xe)
> > }
> > }
> >
> > - ret = xe_bo_apply_to_pinned(xe, &xe->pinned.external_vram,
> > - &xe->pinned.external_vram,
> > + ret = xe_bo_apply_to_pinned(xe, &xe->pinned.external,
> > + &xe->pinned.external_evicted,
> > xe_bo_evict_pinned);
>
> Just one question here. Do you know if this changes the existing
> behaviour during suspend-resume for such buffers? Say we have pinned
> tt
> that was exported to another driver (not dynamic), will this now move
> it
> from tt -> system? If so, is that correct? Does it make sense to skip
> this entirely for suspend-resume for the external pinned case? I'm
> assuming it can't be in VRAM for this case, so seems like we don't
> need
> to do anything here?
Currently xe_bo_evict_pinned() doesn't do anything for !vram bos, and
we currently don't allow pinning bos in VRAM, so yes as a consequence
we could probably remove the above.
I was a bit hesitant to include the introduction of the
external_evicted list in this patch, since that's really a separate
change, so perhaps the best choice here is to remove the introduction
of external_evicted() and we could then remove the suspend / resume
iteration over pinned dma-buf entirely in a separate patch. I was about
to send out a v3 anyway so I'll restrict this patch to only the
"remove" callback addition and then we could add another patch to deal
with the possibly unnecessary iteration at suspend / resume time.
Thanks,
Thomas
>
> Otherwise looks good.
>
> >
> > /*
> > @@ -181,8 +181,8 @@ int xe_bo_restore_user(struct xe_device *xe)
> > return 0;
> >
> > /* Pinned user memory in VRAM should be validated on
> > resume */
> > - ret = xe_bo_apply_to_pinned(xe, &xe->pinned.external_vram,
> > - &xe->pinned.external_vram,
> > + ret = xe_bo_apply_to_pinned(xe, &xe-
> > >pinned.external_evicted,
> > + &xe->pinned.external,
> > xe_bo_restore_pinned);
> >
> > /* Wait for restore to complete */
> > @@ -191,3 +191,57 @@ int xe_bo_restore_user(struct xe_device *xe)
> >
> > return ret;
> > }
> > +
> > +static void xe_bo_pci_dev_remove_pinned(struct xe_device *xe)
> > +{
> > + struct xe_tile *tile;
> > + unsigned int id;
> > +
> > + (void) xe_bo_apply_to_pinned(xe, &xe->pinned.external,
> > + &xe->pinned.external,
> > + xe_bo_dma_unmap_pinned);
> > + for_each_tile(tile, xe, id)
> > + xe_tile_migrate_wait(tile);
> > +
> > + (void) xe_bo_apply_to_pinned(xe, &xe-
> > >pinned.kernel_bo_present,
> > + &xe->pinned.kernel_bo_present,
> > + xe_bo_dma_unmap_pinned);
> > +
> > +}
> > +
> > +
> > +
> > +/**
> > + * xe_bo_pci_dev_remove_all() - Handle bos when the pci_device is
> > about to be removed
> > + * @xe: The xe device.
> > + *
> > + * On pci_device removal we need to drop all dma mappings and move
> > + * the data of exported bos out to system. This includes SVM bos
> > and
> > + * exported dma-buf bos. This is done by evicting all bos, but
> > + * the evict placement in xe_evict_flags() is chosen such that all
> > + * bos except those mentioned are purged, and thus their memory
> > + * is released.
> > + *
> > + * For pinned bos, we're unmapping dma.
> > + */
> > +void xe_bo_pci_dev_remove_all(struct xe_device *xe)
> > +{
> > + unsigned int mem_type;
> > +
> > + /*
> > + * Move pagemap bos and exported dma-buf to system, and
> > + * purge everything else.
> > + */
> > + for (mem_type = XE_PL_VRAM1; mem_type >= XE_PL_TT; --
> > mem_type) {
> > + struct ttm_resource_manager *man =
> > + ttm_manager_type(&xe->ttm, mem_type);
> > +
> > + if (man) {
> > + int ret =
> > ttm_resource_manager_evict_all(&xe->ttm, man);
> > +
> > + drm_WARN_ON(&xe->drm, ret);
> > + }
> > + }
> > +
> > + xe_bo_pci_dev_remove_pinned(xe);
> > +}
> > diff --git a/drivers/gpu/drm/xe/xe_bo_evict.h
> > b/drivers/gpu/drm/xe/xe_bo_evict.h
> > index 746894798852..0f06140163d6 100644
> > --- a/drivers/gpu/drm/xe/xe_bo_evict.h
> > +++ b/drivers/gpu/drm/xe/xe_bo_evict.h
> > @@ -12,4 +12,5 @@ int xe_bo_evict_all(struct xe_device *xe);
> > int xe_bo_restore_kernel(struct xe_device *xe);
> > int xe_bo_restore_user(struct xe_device *xe);
> >
> > +void xe_bo_pci_dev_remove_all(struct xe_device *xe);
> > #endif
> > diff --git a/drivers/gpu/drm/xe/xe_device.c
> > b/drivers/gpu/drm/xe/xe_device.c
> > index b2f656b2a563..739861edcd30 100644
> > --- a/drivers/gpu/drm/xe/xe_device.c
> > +++ b/drivers/gpu/drm/xe/xe_device.c
> > @@ -23,6 +23,7 @@
> > #include "regs/xe_gt_regs.h"
> > #include "regs/xe_regs.h"
> > #include "xe_bo.h"
> > +#include "xe_bo_evict.h"
> > #include "xe_debugfs.h"
> > #include "xe_devcoredump.h"
> > #include "xe_dma_buf.h"
> > @@ -468,8 +469,9 @@ struct xe_device *xe_device_create(struct
> > pci_dev *pdev,
> >
> > spin_lock_init(&xe->pinned.lock);
> > INIT_LIST_HEAD(&xe->pinned.kernel_bo_present);
> > - INIT_LIST_HEAD(&xe->pinned.external_vram);
> > + INIT_LIST_HEAD(&xe->pinned.external);
> > INIT_LIST_HEAD(&xe->pinned.evicted);
> > + INIT_LIST_HEAD(&xe->pinned.external_evicted);
> >
> > xe->preempt_fence_wq = alloc_ordered_workqueue("xe-
> > preempt-fence-wq",
> >
> > WQ_MEM_RECLAIM);
> > @@ -925,6 +927,8 @@ void xe_device_remove(struct xe_device *xe)
> > xe_display_unregister(xe);
> >
> > drm_dev_unplug(&xe->drm);
> > +
> > + xe_bo_pci_dev_remove_all(xe);
> > }
> >
> > void xe_device_shutdown(struct xe_device *xe)
> > diff --git a/drivers/gpu/drm/xe/xe_device_types.h
> > b/drivers/gpu/drm/xe/xe_device_types.h
> > index 4472bf849f2b..aa83ba414301 100644
> > --- a/drivers/gpu/drm/xe/xe_device_types.h
> > +++ b/drivers/gpu/drm/xe/xe_device_types.h
> > @@ -426,8 +426,10 @@ struct xe_device {
> > struct list_head kernel_bo_present;
> > /** @pinned.evicted: pinned BO that have been
> > evicted */
> > struct list_head evicted;
> > - /** @pinned.external_vram: pinned external BO in
> > vram*/
> > - struct list_head external_vram;
> > + /** @pinned.external: pinned external and dma-buf.
> > */
> > + struct list_head external;
> > + /** @pinned.external: evicted pinned external and
> > dma-buf. */
> > + struct list_head external_evicted;
> > } pinned;
> >
> > /** @ufence_wq: user fence wait queue */
>
More information about the Intel-xe
mailing list